STV in practice: Northern Irish local elections
Henry Potts
Raw data from Nick Whyte's webpages for the Alliance Party of Northern
Ireland.
YEAR PARTY % VOTE % CLLR.S # CLLR.S
1973 DUP 4 4.0 21
UUP 41 44.3 233
other U 11 10.1 53
APNI 14 12.0 63
others 16 13.9 73
SDLP 13 15.8 83
1977 DUP 13 14.1 74
UUP 30 33.8 178
other U 9 6.5 34
APNI 14 13.3 70
others 13 10.8 57
SDLP 21 21.5 113
1981 DUP 27 27.0 142
UUP 27 28.9 152
other U 5 4.2 22
APNI 9 7.2 38
others 15 13.1 69
SDLP 18 19.6 103
1985 DUP 24 25.1 142
UUP 29 33.6 190
other U 3 2.5 14
APNI 7 6.0 34
others 6 4.6 26
SDLP 18 17.8 101
SF 12 10.4 59
1989 DUP 18 19.5 110
UUP 31 34.3 194
other U 6 5.7 32
APNI 7 6.7 38
others 6 4.8 27
SDLP 21 21.4 121
SF 11 7.6 43
1993 DUP 17 17.7 103
UUP 29 33.8 197
other U 4 4.8 28
APNI 8 7.6 44
others 7 5.5 32
SDLP 22 21.8 127
SF 12 8.8 51
(There are some interesting trends over time: the introduction of SF had
no effect on the SDLP vote, seeming to come more from the Others.
Interestingly, the small bias towards SDLP under this electoral system
(see below) vanished when SF appeared - I have no idea why that should
be! And why the fall in Alliance support after 1977?)
"Other U" is a catch-all of various Unionist groups; "Others" is a
catch-all for everyone left over. These groupings might see worse
proportionality as they consist of several small parties. OTOH, these
'others' may be local indepedendents with highly concentrated support,
who should do better than proportionality. Unfortunately, the two
'other' categories account for a considerable proportion of the vote
(27% in 1973).
With an average district size of four (which is a complete guess),
Lijphart (1994) would give an effective theshold of 16.25%. By "small
party" below, I shall mean (approx.) less than that figure.
Vs. proportionality difference
did better did worse mean range min. max.
DUP 4 2 +0.7% 1.5 -0.0 +1.5
UUP 6 0 +3.6% 2.9 +1.9 +4.8
other U 1 5 -0.7% 3.3 -2.5 +0.8
APNI 0 6 -1.0% 1.7 -2.0 -0.3
others 0 6 -1.7% 1.0 -2.2 -1.2
SDLP 4 2 +0.8% 3.0 -0.2 +2.8
SF 0 3 -2.7% 1.8 -3.4 -1.6
The mean absolute bias (excluding the two 'other' categories) is 1.7% or
about 10 councillors (in 1993). Overall, the results show admirable
proportionality, but there is a *consistent* pattern of bias. In
particular, bias towards the UUP and against SF.
If the disproportionality comes from the prefential system (i.e.
considering first preference share is a poor measure and the apparent
bias is not a fault of STV), we would expect more centrist parties to do
better and more extremist ones to do worse. (I'm using 'centrist' in a
loose sense given the sectarian nature of politics in NI.)
Alternatively, if the disproportionality comes more from bias against
smaller parties, then big parties should do well and small ones badly.
SF does worst, an extreme and a small party. The UUP does best, the less
extreme of the two main Unionist parties and the biggest party overall.
Crucially, Alliance sees a mild bias against it, suggesting it suffers
from being a small party more than it gains from being centrist. We know
too little to deduce much from the two 'other' categories' results.
Clearly, there is an enourmous amount of geographical heterogeneity in
Northern Ireland, making it a poor example to consider. I hope I haven't
got anything about the parties confused; us mainlanders can find NI
politics very confusing, so please forgive me if I have! I would be
interested in comments from anyone (i.e. Nick) more au fait with NI
political history and how the parties behave under STV.
--
Henry
Henry Potts
In article <memo.9701022...@rosenstiel.compulink.co.uk>, Colin
Rosenstiel <rosen...@cix.co.uk> writes
>In article <fW+J$DACKn...@bondegezou.demon.co.uk>,
>
>> Raw data from Nick Whyte's webpages for the Alliance Party of
>> Vs. proportionality difference
>> did better did worse mean range min. max.
>
>
>This ignores the limitations of the relatively small numbers involved
582 councillors (in 1993) seems a sufficiently large number. Apart from
district sizes being smaller -- does anyone know how big they are in
Northern Irish local elections? -- we have a model close to electing the
House of Commons under the ERS's suggestions.
>and the uneven size of the authorities.
Largely irrelevant -- each district (ward) elects its councillors. As an
illustration of the effect of the elctoral system, it is irrelevant that
those councillors sat in various unequally-sized authorities.
>It looks pretty proportional to me. Look at other PR systems and they
>won't be significantly more precise on this scale.
I am using these results to illustrate STV's degree of proportionality,
not to argue for a different system for Northern Ireland's local
elections. If it had been possible to hold these elections as a single-
district list system, the results would have been somewhat more
proportional. Of course, that would be a silly way to elect local
councils, but when considering electoral systems for national
legislatures, that is a choice open to us. The results are "pretty
proportional", although they could be more proportional. So it depends
how important you consider proportionality how you interpret these
results.
I wanted to illustrate the behaviour of smaller parties under STV, but
the numerous "others" and extreme geographical heterogeneity rather
prevented this.
--
Henry
Nicholas Whyte
In article <fW+J$DACKn...@bondegezou.demon.co.uk>,
he...@bondegezou.demon.co.uk says...
>
>Raw data from Nick Whyte's webpages for the Alliance Party of Northern
>Ireland.
Hah! Someone’s been reading them then.
>(There are some interesting trends over time: the introduction of SF had
>no effect on the SDLP vote, seeming to come more from the Others.
This is because many of the Others were independent Nationalists, some of
whom supported Sinn Fein. Also SF motivated a number of voters who simply
hadn’t bothered voting before.
>Interestingly, the small bias towards SDLP under this electoral system
>(see below) vanished when SF appeared - I have no idea why that should
>be! And why the fall in Alliance support after 1977?)
The small SDLP bias is not really significant given the effects of sampling
(of which more below). The decline in the Alliance vote is generally
attributed to the polarising effect of the 1980-81 hunger strikes and the
contemporary rise of Paisleyism - the DUP had its best election results at
that time as well.
>"Other U" is a catch-all of various Unionist groups; "Others" is a
>catch-all for everyone left over. These groupings might see worse
>proportionality as they consist of several small parties. OTOH, these
>'others' may be local indepedendents with highly concentrated support,
>who should do better than proportionality. Unfortunately, the two
>'other' categories account for a considerable proportion of the vote
>(27% in 1973).
The largest part of this was the Vanguard Unionist Party which disintegrated
in 1975.
>With an average district size of four (which is a complete guess),
>Lijphart (1994) would give an effective theshold of 16.25%. By "small
>party" below, I shall mean (approx.) less than that figure.
The average district size was actually just under six (at present we have
101 electoral areas for 582 seats; average district size has increased
slightly but not much in the two revisions since 1973).
Isn’t the "effective threshold" in fact empirically calculated, as the
average between the lowest % at which one gets elected and the highest % at
which one fails to get elected? There were several cases in the last local
elections of people getting in on transfers with less than 5% of first
preferences, in some cases because they were running mates of a prominent
personality who got loads of first prefs, and in at least two cases because
they were members of a "centre" party (Alliance in Banbridge, Workers Party
in Craigavon) who benefited from undernomination by nationalist parties.
Similarly there were a couple of cases of people and parties who came close
to the quota in first preferences and got no transfers (Alliance in the
Downshire area of Lisburn comes to mind, we got 13% in a five-seater and
still didn't get in - see below for more on how the system is more likely to
work against us than for us). So I’d put the "effective threshold" a lot
lower than 16%.
>The mean absolute bias (excluding the two 'other' categories) is 1.7% or
>about 10 councillors (in 1993). Overall, the results show admirable
>proportionality, but there is a *consistent* pattern of bias. In
>particular, bias towards the UUP and against SF.
>
>If the disproportionality comes from the prefential system (i.e.
>considering first preference share is a poor measure and the apparent
>bias is not a fault of STV), we would expect more centrist parties to do
>better and more extremist ones to do worse. (I'm using 'centrist' in a
>loose sense given the sectarian nature of politics in NI.)
>Alternatively, if the disproportionality comes more from bias against
>smaller parties, then big parties should do well and small ones badly.
>SF does worst, an extreme and a small party. The UUP does best, the less
>extreme of the two main Unionist parties and the biggest party overall.
>Crucially, Alliance sees a mild bias against it, suggesting it suffers
>from being a small party more than it gains from being centrist. We know
>too little to deduce much from the two 'other' categories' results.
>Clearly, there is an enourmous amount of geographical heterogeneity in
>Northern Ireland, making it a poor example to consider. I hope I haven't
>got anything about the parties confused; us mainlanders can find NI
>politics very confusing, so please forgive me if I have! I would be
>interested in comments from anyone (i.e. Nick) more au fait with NI
>political history and how the parties behave under STV.
I am inclined to blame the voters rather than the system(!) STV in local
elections here has an excellent record of getting smaller parties some
representation (if you look at my /councils.html page on the Alliance site
you will count around a dozen groups with at least one councillor). The
biggest reason for the bias towards the UUP and against SF and Alliance is
the preferential voting system: in most areas where SF stand, they don’t get
many transfers because they are an extreme party; and (surprisingly) this is
also true of Alliance. There are at present 45 Alliance councillors
representing 33 areas, and fewer than half of them had *any* SDLP or Sinn
Fein candidates running against them at the last election (only two actually
share their constituencies with a Sinn Fein councillor). In most areas where
Alliance is strong, the "centre" party is the UUP, and the preferential
system accordingly benefits them (as in Downshire, mentioned above). I did a
count of the 101 areas some time back and if you arrange the votes in a
spectrum in each one the UUP comes out as the "centre-most" party more than
half the time. (see http://www.niweb.com/nielection/table.html for partial
results of this)
There is another factor as well (as Colin points out): the fact that council
areas vary in size. In Belfast we have about 4,000 voters per councillor; in
some of the rural areas it’s less than a thousand. Of all the parties the
UUP is the most rurally based, so the fact that there are more councillors
per voter in its stronger areas means that "proportionality" gets distorted.
This applies conversely to Sinn Fein, of course; around a quarter to a third
of its total vote across Northern Ireland in May 1993 was concentrated into
three electoral areas in Belfast where there was an exceptionally high
turnout (like 70% compared with an average of 50% for the rest of the city)
so effective a Sinn Fein vote elects fewer councillors than a vote for any
other party if you look at it in a Norrthern Ireland wide context (which may
note be the most appropriate context for elections to 26 different local
councils).
Incidentally, I’d like to take this opportunity to wish you all a fruitful
1997. I have just got a twelve-month contract as a trainer of political
parties in Bosnia, starting next week, so I don’t know how much I will be
contributing to uk.politics.electoral this year, though I suspect that it
will be less than last year! STV consultancy will be available at my usual
rates...
All the best everyone,
Nicholas Whyte
Henry Potts
In article <5aip62$6...@fu-berlin.de>, Nicholas Whyte <nwh...@unite.net>
writes
>>Ireland.
>
They surely deserve their description as the best political party's
pages on the WWW.
>>With an average district size of four (which is a complete guess),
>>Lijphart (1994) would give an effective theshold of 16.25%. By "small
>>party" below, I shall mean (approx.) less than that figure.
>
>The average district size was actually just under six (at present we have
>101 electoral areas for 582 seats; average district size has increased
>slightly but not much in the two revisions since 1973).
>
>Isn't the "effective threshold" in fact empirically calculated, as the
>average between the lowest % at which one gets elected and the highest
>% at which one fails to get elected? There were several cases in the
>last local elections of people getting in on transfers with less than
>5% of first preferences, [...] Similarly there were a couple of cases
>of people and parties who came close to the quota in first preferences
>and got no transfers (Alliance in the Downshire area of Lisburn comes
>to mind, we got 13% in a five-seater and still didn't get in [...]). So
>I’d put the "effective threshold" a lot lower than 16%.
For an average district size of 5.76, Lijphart would give an effective
threshold of 11.7%. Your figures would suggest an empirical figure of
around 9%, which is pretty close. Thanks for your other comments.
[...]
Congratulations on your job in Bosnia. Perhaps we'll see you on
bosnia.politics.electoral? :)
--
Henry
Colin Rosenstiel
In article <+1ix3kA2...@bondegezou.demon.co.uk>,
he...@bondegezou.demon.co.uk (Henry Potts) wrote:
> >This ignores the limitations of the relatively small numbers involved
>
> 582 councillors (in 1993) seems a sufficiently large number. Apart from
> district sizes being smaller -- does anyone know how big they are in
> Northern Irish local elections? -- we have a model close to electing the
> House of Commons under the ERS's suggestions.
>
> >and the uneven size of the authorities.
>
> Largely irrelevant -- each district (ward) elects its councillors. As an
> illustration of the effect of the elctoral system, it is irrelevant that
> those councillors sat in various unequally-sized authorities.
The point I was making was that calculating global proportionality where
the units vary in size does have limitations. I was saying that the
results are as proportional as reasonable for a representative system.
The only significant variation from strict proportionality is for Sinn
Fein. This may be a result of the literal meaning of their name. There are
benefits under STV to parties that attract votes from supporters of other
parties. I would argue that that was a positive attribute of an electoral
system which is not to produce pure proportionality but to produce
political solutions, which require representatives to be willing to work
together.
Cllr. Colin Rosenstiel
Electoral Reform Society Council member